Professional Documents
Culture Documents
Abstract: Recent research on ultrahigh-performance concrete (UHPC) for connection elements in highway bridges indicates an
improvement in the performance of the overall bridge superstructure. However, there is limited information on the effect of the material
and bond strength of the UHPC on the load transfer mechanism between adjacent box girders. A three-dimensional finite-element (FE)
model was developed to study the interface between UHPC and high-strength concrete (HSC) highway bridge connections. The UHPC-
HSC interface was modeled using traction-separation, damage initiation criteria, and damage evolution while taking into account adhe-
sion, friction, and the nonlinear material behavior. Material properties were determined directly from laboratory testing, and the interface
parameters were identified through model calibration using direct tension test results and friction coefficients reported in the previous
work. The model was validated by simulating the laboratory tests conducted at the Federal Highway Administration (FHWA) Turner-
Fairbank Highway Research Center (TFHRC), and good agreement between numerical and experimental results was obtained. Finally,
numerical simulations of two adjacent box-girder models using the proposed interface parameters and other interface models from past
research were compared, and the results highlight the importance of using a high-fidelity model to accurately represent the system behav-
ior. DOI: 10.1061/(ASCE)BE.1943-5592.0001070. © 2017 American Society of Civil Engineers.
Author keywords: Adhesion; Ultrahigh-performance concrete (UHPC); High-strength concrete (HSC); ASTM C1404/C1404M-98; Bond
strength; Friction coefficient; Interfacial properties; Bridge connections; Adjacent box-beam; Finite-element (FE) method.
Fig. 1. (a) FHWA adjacent box-beam connection details using UHPC and transverse SRBs; (b) SRB details
concrete. In a later study, the same FHWA shear key design was modeling showed that only cracking and crushing in the grout/con-
used in the Sollars Road Bridge over Lees Creek in Fayette County, crete occurred, and the model was unable to capture bond failure at
Ohio, to study its performance under field conditions (Steinberg the joint interface, which was the primary mode of failure observed
et al. 2015). The bridge consisted of seven precast prestressed box during testing. Badwan and Liang (2007) presented the results of
beams and had a span of 18.6 m. Similar to the FHWA specimen, live-load testing of precast posttensioned concrete multibeam decks
12.7-mm-diameter steel SRBs were embedded 457 mm into the and associated linear FE analysis. In the FE model, the authors
beams and 127 mm into the UHPC shear keys and were spaced at assumed that there was no adhesion or friction at the interfaces
203 mm in the longitudinal direction. To evaluate the field perform- between joints and beams. A comparison of the experimental and
ance of the bridge, the beams and shear keys were instrumented to the FE simulation results revealed differences in the strain data with
measure beam displacement, strain, and temperature under truck the errors ranging from 18 to 30%. Fu et al. (2011) studied the effect
and environmental loads (Steinberg et al. 2015). of TPT on the behavior of adjacent box beams under a truck load
Research on UHPC for connection elements in bridges indicated and used experimental data to calibrate a FE model of the bridge. In
an improvement in the performance of the overall bridge super- accordance with the American Concrete Institute (ACI) code, the
structure. The improvement in performance can be attributed to the authors assumed a friction coefficient of 0.6 and no adhesion at the
higher strength of the UHPC relative to traditional grouting materi- interface between the shear keys and beams. The model was based
als and the higher bond strength between the UHPC and other on the following assumptions: (1) pin-roller boundary conditions,
bridge components due to the high adhesion of the UHPC material (2) full bond between beams and concrete overlay, and (3) linear
(Hussein et al. 2016). Although the significance of adhesion on the elastic material behavior for all components. To calibrate the FE
bond strength of UHPC has been demonstrated in laboratory test model to the experimental results, the researchers added a spring at
specimens, there is limited information on the effect of these prop- the beam ends and varied the stiffness until agreement between the
erties of the UHPC on the load transfer mechanism between adja- model and actual bridge behavior was obtained. Another FE model
cent box beams. An economical approach to analyzing the UHPC was developed by Ulku et al. (2010) to study the TPT procedure for
joint is through finite-element (FE) modeling. However, such analy- an adjacent box-beam bridge and to calculate the stresses along the
sis requires a detailed description of the mechanical and material longitudinal joints between beams. The authors assumed a full bond
properties of the interface components as well as of the interface between the joints and beams, and the behavior at the interface
itself so that the interaction between all components at the interface between components was not considered in the model. Porter et al.
can be accurately represented. (2011) evaluated five deck panel connections experimentally and
numerically to determine the cracking and ultimate flexural strength
Background of the joints. A cohesive zone material was used to model the inter-
face behavior between the joint and deck panel surfaces. However,
Past research on highway bridges has shown that the strength at the the contact strength of the cohesive material was assumed to be the
interface between components plays an important role in the load same because the maximum tensile strength of the concrete and
transfer between bridge members. In several experimental studies, friction at the interface was not considered. Grace et al. (2008,
the effect of different parameters on the concrete-to-concrete and 2013) developed a three-dimensional (3D) FE model to investigate
UHPC-to-concrete interface strength was investigated. In addition, the effect of TPT and the number of transverse diaphragms on the
several FE models have been calibrated using small-scale or full- behavior of adjacent box-beam bridges using unbonded carbon fiber
scale experimental results to predict the overall bridge behavior. composite cables. Nonlinear analysis was used to model the con-
Issa et al. (2003) studied the performance of four different grout crete behavior, and the interface between shear keys and the beams
materials for the joints between concrete components using vertical depended only on the friction coefficient. The authors neglected the
shear, direct tension, and flexural tests. A FE program was used to effect of adhesion between shear keys and beams. Dias-da-Costa
model the vertical shear test. However, the authors did not provide et al. (2012) calibrated a numerical model with experimental data
any details regarding the interface between the joint and concrete using a push-off test composed of two identical L-shaped halves.
components as represented in the model. Results of the FE The longitudinal shear strength between two normal concrete layers
δnf Separation
model to capture the interface behavior between concrete elements
cast at different times.
Fig. 2. Typical traction-separation response with exponential damage
evolution
FE Analysis
!2 !2 !2
htn i hts i htt i
Material Modeling þ þ ¼1 (2)
tno tso tto
In the present study, all FE analyses were performed using the
commercial software package Abaqus 6.12-3. There are two con-
where tno , tso , and tto = maximum values of the contact stresses.
crete constitutive models available in Abaqus, namely the con-
During the FE simulation, damage is initiated only when the value
crete smeared cracking (CSC) and concrete damaged plasticity
from Eq. (2) is equal to 1. Once damage initiation has occurred,
(CDP) models. Chen and Graybeal (2010, 2012a, b) recom-
damage evaluation is determined based on the total fracture energy
mended using the CDP model for UHPC rather than the CSC
(Gc) or the total plastic displacement at failure (d fn Þ, and damage
model due to a convergence problem and a lack of agreement
evolution softening response can be either linear, exponential, or
with the experimental results in the nonlinear stage. In this study,
user defined. It should be pointed out that some convergence diffi-
the CDP model was used to model concrete constitutive behav-
culties may occur when using traction-separation behavior.
iors. The nonlinear behavior of concrete was represented by iso-
However, these difficulties may be overcome by defining the vis-
tropic damage elasticity combined with isotropic tensile and com-
cous regularization of the constitutive equations.
pressive plasticity. A softening stress-strain relationship was used
to represent tensile microcracks, whereas the compressive plastic
response followed by strain softening was used beyond the ulti- FE Models
mate compressive strength. In Abaqus, a viscoplastic regulariza-
tion can be used to improve the convergence rate in the softening Three different 3D FE models were used in the present study. The
stage, using the ultimate tensile plastic strain and the maximum first model was based on the direct tension test (ASTM C1404/
tensile stress. The CDP model requires five additional parameters C1404M-98) (ASTM 2003) and was used along with experimental
including flow potential eccentricity (e ), a viscosity parameter data to identify the interface parameters between UHPC and HSC.
that defines viscoplastic regularization ( m ), the ratio of the second The second model was based on the slant-shear test (ASTM C882/
stress invariant on the tensile meridian to that on the compressive C882-13a) (ASTM 2013) and was used to validate the interface pa-
meridian such that the maximum principal stress is negative (Kc), rameters along with the friction coefficients, which were back-
the ratio of the initial equibiaxial compressive yield stress to the calculated in previous work using the Mohr-Coulomb theory and
initial uniaxial compressive yield stress (s b0/s c0), and the dila- direct tension test results (Hussein et al. 2016). The third model con-
tion angle in degrees ( c ). sisted of two adjacent box beams connected by a UHPC joint and
transverse SRBs and was used to validate the interface model via
Interface Modeling the global beam behavior. The slant-shear and adjacent box-beam
models were based on tests performed at FHWA-TFHRC, and data
To model the concrete-to-concrete interface, a traction-separation from these tests were used in the validation.
model was used that requires material surface properties. The
traction-separation constitutive model offers a method to model
Direct Tension Model
adhesive/cohesive connection by assuming zero thickness of the
interface. The traction-separation behaviors are linear elastic The direct tension test method (ASTM C1404/C1404M-98)
traction-separation, DIC, and damage evolution models. The (ASTM 2003) was used to measure the adhesion value at the
linear elastic traction-separation parameters, Kn, Ks, and Kr, are UHPC-HSC interface by applying a direct tensile force to a UHPC-
the normal and tangential stiffness components that relate to the HSC specimen until failure took place (Hussein et al. 2016).
normal and shear separation across the interface before the ini- Fig. 3(a) shows the cross section, test apparatus, and load configura-
tiation of damage. The contact stresses at the interface are the tion. Three different degrees of surface roughness for the HSC sub-
normal contact stress in the pure normal mode (tn), the shear strate were considered, namely smooth, midrough, and rough. The
contact stress along the first shear direction (ts), and the shear smooth surface was the flat cut surface specified in ASTM C1404/
contact stress along the second shear direction (tt), as shown in C1404M-98 (ASTM 2003), the midrough surface was a sandblasted
Fig. 2; DIC occurs when the following quadratic stress-based surface specified by the Ohio DOT Construction Specification
damage criterion for a cohesive surface is satisfied: 706.05 (ODOT 2013), and the rough surface was the exposed
Fig. 3. (Color) Direct tension test (ASTM C1404/C1404M-98): (a) cross section of mold assembly; (b) FE model
180
HSC
160
UHPC
140
120
Stress (MPa)
100
80
60
40
20
0
0 0.005 0.01 0.015 0.02
Strain
aggregate surface recommended for UHPC joints by Russell and cylinders and steel components were modeled with 8-node brick
Graybeal (2013) and Graybeal (2014). The average tensile stresses elements with a maximum mesh size for the cylinders of 5 mm.
and the mode of failure for the smooth, midrough, and rough surfa- Boundary conditions were applied to the bottom of the steel bolt to
ces were measured to be 3.02 MPa at the bond line, 5.01 MPa at the restrain its vertical and horizontal movement, and the load was
transition zone, and 5.63 MPa at the concrete substrate, respectively applied to the other end. The interface between the UHPC and HSC
(Hussein et al. 2016). cylinders was modeled using the traction-separation constitutive
The FE model of the direct tension test consisted of five compo- model. The elastic behavior of the HSC, UHPC, and steel were
nents, the steel bolt, the pipe cap, the steel nipple, the HSC cylinder, defined by their modulus of elasticity and Poisson’s ratio. It should
and the UHPC cylinder, as shown in Fig. 3(b). The concrete be noted that the UHPC used in the present study is the Ductal
steel fibers once the UHPC has cracked (Chen and Graybeal 2010, a maximum mesh size of 5 mm. Boundary conditions were applied
2012a, b). The tensile stress at failure was determined from indirect to the bottom surface of the HSC side to restrain the vertical and
tensile test (ASTM C496/C496M-11) (ASTM 2004) data. The CDP horizontal movement, and the load was applied to the top surface of
model parameters for the UHPC and HSC are listed in Table 1. For the UHPC side. The UHPC and HSC used in the FHWA-TFHRC
the steel components, the Poisson’s ratio and modulus of elasticity research had similar mix designs to those used in Hussein et al.
were taken to be 0.3 and 200 GPa, respectively. To establish the (2016), and the properties in Table 1 and Fig. 4 were used to repre-
interface parameters between the UHPC and HSC materials in the sent the elastic and inelastic behavior of the UHPC and HSC materi-
traction-separation model, the normal and tangential stiffness com- als. The interface between the UHPC and HSC cylinders was mod-
ponents (Kn, Ks, and Kr) and the total plastic displacement were cali- eled using the traction-separation constitutive model, with normal
brated by fitting the load-displacement curves from the direct ten- and tangential behaviors represented using hard contact and fric-
sion models to the experimental data. The maximum normal and tion, respectively. Model validation was achieved by loading the
tangential strengths (tno , tso , and tto ) were assumed to be the same in UHPC-HSC to failure and comparing the results with the FHWA-
the normal and tangential directions, and values of 3.02, 5.01, and TFHRC test data.
5.63 MPa were used for the smooth, midrough, and rough surfaces,
respectively. These values were determined via the direct tension Adjacent Box-Beam Model
test (ASTM C1404/C1404M-98) (ASTM 2003) and were previ-
The adjacent box-beam model was based on a load-test configura-
ously reported by Hussein et al. (2016).
tion used at the FHWA-TFHRC (Yuan and Graybeal 2016). The
box beams were connected with a UHPC joint and transverse steel
Slant-Shear Model SRBs. The two adjacent box-beam cross sections, the joint connec-
The slant-shear tests (ASTM C882/C882M-13a) (ASTM 2013) tion details, loading configuration, and instrumentation plan are
simulated in the present work were conducted at the FHWA- depicted in Fig. 5. The beams were 15.25 m long, 0.84 m deep, and
TFHRC. The test results were used to back-calculate the friction 0.91 m wide, and were prestressed using 15.24-mm-diameter low-
relaxation prestressing strands, all stressed to 195.7 kN. Each beam
contained 27 strands spaced in a grid pattern with 50.8-mm spacing,
Table 1. UHPC and HSC Properties Used in FE Modeling
and 9 of the strands were debonded 1.8 m from both ends of the
Property UHPC HSC beam. The beam material was HSC with a compressive strength of
a 67.8 MPa at the time of the test. The longitudinal joint between the
Modulus of elasticity (GPa) 53 41a
adjacent beams was filled with the Ductal JS1000 brand UHPC
Poisson’s ratio 0.19a 0.17a
manufactured by Lafarge North America. Details of the joint are
Compressive strength (MPa) 158.58a 75a
shown in Fig. 5(d). TPT was not used to take full advantage of the
Maximum tensile stress (MPa) 15.9b 4.8a
UHPC material. However, the two adjacent box beams had 12.7-mm-
Fracture energy [Gf (N/m)] 87,559b 120c
diameter steel SRBs, Grade 60, installed transversely in the UHPC
Dilation angle ( c ) 15°b 36°c
joints to help with load transfer between beams. The SRBs were em-
Eccentricity (e ) 0.1b 0.1c
bedded 457 and 140 mm in the beam and shear key, respectively, as
s b0/s c0 1.16b 1.16c
shown in Fig. 5(c), and were spaced at 203 mm in the longitudinal
Kc 2/3b 2/3c
direction. Also, the exposed aggregate surface (rough surface) was
Viscosity parameter ( m ) 0.0b 0.001c
used to enhance the bond strength between the UHPC and HSC ma-
a
Test. terial. On one end of the beams [Fig. 5(e), Support A-A], a 305
b
Chen and Graybeal (2010, 2012a, b). 762 mm steel plate with 50.8-mm thickness was connected to
c
Abaqus. another steel plate (305 457 mm with 50.8-mm thickness), which
Table 2. Friction Coefficients Back-Calculated Using the Results of Slant-Shear and Direct Tension Tests
Fig. 5. (a) Cross section at Support A-A; (b) cross section at Support B-B; (c) loading configuration and instrumentation plan; (d) joint configuration;
(e) plane view of adjacent box beams
Fig. 6. (Color) (a) Test setup; (b) joint-SRBs before casting with UHPC; (c) Support A-A
120 South
North
100
Downloaded from ascelibrary.org by University of California, San Diego on 06/19/17. Copyright ASCE. For personal use only; all rights reserved.
80
Load kip
60
40
20
0
0 50 100 150 200
Time
Fig. 8. (Color) FE mesh of the full model of two adjacent box beams
were prestressed to 1,392.7 MPa, and elastic shortening was beams at midspan, resulting in total loads of 22.2 and 422.6 kN on
accounted for in the simulation. To capture the behavior of the the two beams. The difference in beam response obtained from the
beams with the constraining mechanisms at each end, the mecha- first and second loading steps was compared with the experimental
nism was modeled in the FE program. The tie bars at each end were result.
stressed to 220 MPa. Hard contact and a friction coefficient of 1.0
accounted for the normal and tangential behaviors at the interface
between the beams and the steel support plates. The elastic behavior Results and Discussion
of all steel components was defined by the modulus of elasticity and
Poisson’s ratio. The properties shown in Table 1, and those derived
from Fig. 4, were used to represent the elastic and inelastic behavior Direct Tension Model
of the UHPC and HSC materials. Also, material densities of 25, 24, Normal and tangential stiffness components for the linear elastic
and 77 kN/m3 for UHPC, HSC, and steel, respectively, were used to traction-separation model and the total plastic displacement for
calculate the deformation of these materials under the gravity load. damage evolution were obtained by calibrating the direct tension
The interface between the UHPC joint and HSC beams was mod- model so that the load-displacement curves matched the experimen-
eled using the traction-separation constitutive model, and the rough tal results for the three surface roughness types. A summary of the
calibrated interface model parameters between the UHPC and HSC
Table 3. Mechanical Properties of Interface for the different roughness types is provided in Table 3. A compari-
Property Smooth Midrough Rough son of load-deflection curves resulting from the calibrated FE mod-
els and the experiments is shown in Fig. 9. Generally, good agree-
3
Knn (N/mm ) 1,358 1,358 1,358 ment between the numerical and experimental data was observed.
Kss and Ktt (N/mm3) 20,358 20,358 20,358 Another way of validating the interface behavior is to compare
tno , tso , and tto (MPa) 3.02 5.01 5.63 the failure modes from the FE simulation with those observed dur-
Total/plastic displacement (mm) 0.018 0.117 0.241 ing testing. According to Hussein et al. (2016), tensile failure for the
Stabilization 0.001 0.001 0.001 UHPC-HSC test specimen with a smooth interface surface took
30
Sm-EXP
Mr-EXP
Ro-EXP
25 Sm-FEM
Mr-FEM
Ro-FEM
20
Load (kN)
15
10
0
0 0.5 1 1.5 2 2.5 3 3.5
Displacement (mm)
Fig. 9. Experimental (EXP) and finite-element model (FEM) load-displacement curves for smooth (Sm), midrough (Mr), and rough (Ro) surfaces
Fig. 10. (Color) Failure modes for (a) smooth, (b) midrough, and (c) rough surfaces from FE models and failure modes for (d) smooth, (e) midrough,
and (f) rough surfaces from experimental tests
Fig. 11. (Color) EPT and failure modes of FE slant-shear model with one deformation scale factor for (a) smooth, (b) midrough, and (c) rough
UHPC-HSC interface surfaces
place at the bond line. For the midrough surface, the tensile failure with some damage to the HSC. Again, the result is consistent with
primarily occurred in the transition zone adjacent to the bond line, the failure mode observed during testing [Fig. 10(e)]. For the rough
whereas failure for the rough surface occurred mainly in the HSC surface, the EPT value was even higher than that for the midrough
substrate. The results of the FE simulations using the calibrated surface and occurred over a larger area, as shown Fig. 10(c). The
interface behavior showed good agreement with the experimental failure in the test for the rough surface occurred mainly in the HSC
results in terms of the failure modes. Figs. 10(a–c) represent the elements with some damage at the interface between UHPC and
simulated equivalent plastic strain (EPT) in the HSC for different HSC and was consistent with the observed failure mode during test-
types of interface surface roughness. For the smooth surface shown ing, as shown in Fig. 10(f).
in Fig. 10(a), the existence of EPT was limited to a small area and
indicated that bond failure occurred at the interface between the
Slant-Shear Model
materials, with very little damage occurring in the HSC. The result
is consistent with the actual failure mode observed during testing, The slant-shear models of UHPC-HSC specimens with different
as shown in Fig. 10(d). For the model with the midrough surface, as degrees of surface roughness were used to validate the interface pa-
shown in Fig. 10(b), the EPT in the HSC is higher and occurs over a rameters determined in the previous section along with the friction
slightly larger area, indicating that failure occurred at the bond line coefficient calculated in previous work (Tables 2 and 3). To do so,
UHPC bond resisted the tangential slip. Once the adhesive strength used by some researchers included the standard assumption of sim-
of the UHPC bond reaches the maximum value and damage was ini- ple supports with additional calibration by adding spring stiffening
tiated, the contribution of adhesion to the shear stress was reduced, in the longitudinal and transverse directions to create a good match
and the friction model became active. Once the ultimate adhesive between the measured and calculated deflection (Fu et al. 2011;
strength was reached, only the friction contributed to the shear re- Steinberg et al. 2013). In this study, the special mechanism used to
sistance on the interface surface. For the UHPC-HSC with the rough control the degree of rotational constraint at each end was modeled
surface, Hussein et al. (2016) assumed the cohesion value of the in the FE program along with using a multistep analysis that repre-
HSC from the direct tension test to be the minimum value of adhe- sents the sequence in which the two adjacent girders were supported
sion at the material interface and then treated the resulting friction and loaded during construction. The boundary conditions were
coefficient from the back-calculation as a maximum value. This applied to the bottom of the steel plates by restraining the horizontal
approach was validated in the slant-shear simulation with the rough and vertical movements, and frictional contact between the top of
surface as failure took place in the HSC, as shown in Fig. 11(c). the steel plates and the bottom of girder was also simulated.
These modes of failure obtained via FE simulation were consistent To calibrate and validate the adjacent box-beam model, two load
with those observed during FHWA-TFHRC testing (Table 4). cases from the FHWA-TFHRC testing were selected based on hav-
ing the highest deflection at the midspan of the beams during the ex-
Adjacent Box-Beam Model perimental test. The first load case was 22.2 and 422.6 kN on north
and south beams, respectively, whereas the second load case
Model Calibration switched the load locations. Model calibration resulted in values of
The results of the slant-shear simulations in the previous section 4 and 35 GPa for the modulus of the north and south beams, respec-
demonstrated that the interface parameters and friction coefficients tively, and a comparison of the simulated and experimental vertical
for the three types of surface roughness may be used to simulate the deflections at midspan is shown in Fig. 13(a). The second load case
interface behavior in UHPC-HSC connections. In this section, the was then used to check the calibrated model, and the vertical deflec-
tions at midspan are shown in Fig. 13(b). Good agreement between
Table 4. Results of FE Slant-Shear Model the simulated and experimental results for both load cases was
Roughness Experimental average Failure FE model obtained.
degree P (kN) experiment P (kN) Another method that was used to validate the adjacent box-beam
model was a comparison of the beam reactions. The reactions were
Smooth 75.86 Bond 75.713 measured during the laboratory testing using load cells placed
Midrough 252.2 Bond 227.31 between the beams and supports (Fig. 5). There was a single load
Rough 620.4 Concrete 602.75 cell placed beneath each beam at one end of the test specimen and
7.2 7.2
EXP North beam LVDT1
FEM (E=35GPa) South beam LVDT4
7 FEM (E=37GPa) 7
FEM (E=40GPa)
6.8 6.8
Displacement mm
Displacement mm
6.6 6.6
6.4 6.4
6.2 6.2
6 6
5.8 5.8
1 2 3 4 0 50 100 150
LVDT Time
(a) (b)
Fig. 12. Deflection at LVDT locations along midspan with loads of (a) 22.2 kN north and 422.6 kN south beams, and (b) 422.6 kN north and 22.2 kN
south beams (Note: EXP = experimental; FEM = finite-element model)
6.8 6.8
Displacement mm
Displacement mm
6.6 6.6
6.4 6.4
6.2 6.2
6
Downloaded from ascelibrary.org by University of California, San Diego on 06/19/17. Copyright ASCE. For personal use only; all rights reserved.
5.8 5.8
1 2 3 4 1 2 3 4
LVDT LVDT
(a) (b)
Fig. 13. Deflection at LVDT locations along midspan for (a) Load Case 1 and (b) Load Case 2, comparing experimental and simulation values
300 150
EXP EXP
250 FEM 130
FEM
110
200 90
Reaction (kN)
Reaction (kN)
150 70
50
100
30
50 10
-10
0
-30
-50 -50
1 2 1 2 3 4
North beam South beam
North beam South beam
(a) (b)
Fig. 14. Reactions at (a) Support A-A and (b) Support B-B for loads of 22.2 and 422.6 kN on the north and south beams
two load cells placed beneath each beam at the opposite end. The specified in ACI (2011). The ACI model uses a friction coefficient
same conditions were represented in the FE model, as shown in of 0.8, hard contact at the interface between the UHPC-HSC surfa-
Fig. 8. The reactions measured by each load cell for the first load ces, and no adhesion (assumed to be cracked). The second model
case (22.2 and 422.6 kN on north and south beams) are presented in was adopted from Chen and Graybeal (2010, 2012a), who modeled
Fig. 14 along with the simulated values, and good agreement is the interface between UHPC components using hard contact and an
observed at both ends of the beams. Furthermore, the total reaction assumed friction coefficient ranging from 0.1 to 0.3. Although the
at the supports from both the experiments and the simulations was model was UHPC to UHPC, adhesion at the interface was neglected
determined to be the same, that is, 446 kN. The last data point used and the surface interaction was based only on friction, making the
to validate the adjacent box-beam model was the joint opening at results relevant to the present comparison. The third model selected
midspan. The opening reported by FHWA-TFHRC was 0.021 mm, for comparison is based on Steinberg et al. (2013). In the model, a
whereas the opening determined using the FE model was 0.025 mm, friction coefficient of 0.8 is used along with a maximum shear re-
showing good agreement between the model and the experimental sistance at the interface of 5.5 MPa. The interface surface is
measurement. assumed to have a contact stiffness of 13.6 MPa/mm with a limit on
the penetration between the UHPC and HSC components.
UHPC-HSC Interface Model Comparison The ACI (2011), Chen and Graybeal (2010, 2012a), and
It has been shown that the proposed model is capable of represent- Steinberg et al. (2013) models, along with the model proposed in
ing the interface behavior resulting from the combination of adhe- this work, were used to simulate the response of the two adjacent
sion and friction forces in UHPC-HSC connections with smooth, box beams from the FHWA-TFHRC study for the load case having
midrough, and rough surfaces. As discussed previously, past efforts 22.2 and 422.6 kN on the north and south beams, respectively. The
to model the UHPC-HSC interface have used a variety of simulated vertical deflections at midspan of the beams, along with
approaches, and a comparison of the proposed model with those the experimental data from the FHWA-TFHRC tests, are shown in
from past research may be of interest to the reader. The first model Fig. 15 for comparison. The results show that all of the models are
selected for comparison is based on the interface assumptions capable of capturing the general trend in the vertical deflection
6.8
Displacement (mm)
6.6
Downloaded from ascelibrary.org by University of California, San Diego on 06/19/17. Copyright ASCE. For personal use only; all rights reserved.
6.4
6.2
5.8
1 2 3 4
LVDT
Fig. 15. Deflection at LVDT locations along midspan of the beams for different interface models compared with experimental values
observed during testing, but that the model proposed in the present 3. From the adjacent box-girder model, good agreement between
study fits the experimental data the closest. The results demonstrate the simulated and experimental results for both load cases was
the importance of considering friction and adhesion at the interface obtained. This was achieved by considering the special mecha-
between the UHPC and HSC materials and using an appropriate nism for controlling the degree of rotational constraint in the
model for capturing the interface and material behavior. FE model and performing a multistep analysis that represents
the sequence in which the two adjacent girders were loaded
during construction.
Conclusion 4. The interface model from the proposed work, and those from
past research, were used to simulate the response of the two ad-
Recent application of UHPC for bridge connections has resulted in
jacent box girders. The results show that all of the models are
an improvement in the overall bridge performance, which can be
capable of capturing the general trend in the vertical deflection
attributed to the higher strength of the UHPC and the higher bond
that was observed during testing, but that the proposed model
strength between UHPC and the adjacent elements, relative to tradi-
fits the experimental data the closest. The results highlight the
tional grouting materials. However, there has been limited informa-
importance of considering friction and adhesion at the interface
tion on the role of UHPC’s enhanced material and mechanical prop-
between the UHPC and HSC materials and of using a high-
erties on the load transfer mechanism between bridge members. In
fidelity model to represent the system behavior.
the present study, a detailed 3D FE model was developed in which
both the material and interface properties of the connection ele-
ments were considered. Both friction and adhesion at the material
interface were accounted for and the nonlinear material and inter- Acknowledgments
face behaviors were represented. Material properties were deter-
mined directly from laboratory testing, whereas the interface pa- The authors thank the Federal Highway Administration (FHWA)
rameters were identified through model calibration using the results Turner-Fairbank Highway Research Center (TFHRC) staff, who
of direct tension tests and friction coefficients reported in the previ- played a vital part in this research, including Dr. Benjamin Graybeal
ous work. The model was validated via comparison with laboratory and Dr. JiQiu Yuan.
testing conducted at the FHWA-TFHRC. The following conclu-
sions can be drawn based on the results:
1. The direct tension model using the calibrated interface behavior
References
showed good agreement with the experimental results in terms
AASHTO. (2012). LRFD Bridge design specifications, 6th Ed.,
of deflections and the failure modes for all types of surface Washington, DC.
roughness. Abaqus 6.12-3 [Computer software]. Dassault Systèmes SIMULIA,
2. The slant-shear model produced an ultimate load and failure Providence, RI.
mode close to the experimental results for the three types of ACI (American Concrete Institute). (2011). “Building code requirements
surface roughness, highlighting the importance of simulating for structural concrete and commentary.” ACI 318R-11, Farmington
the interface behavior in the UHPC-HSC connection. Hills, MI.
Chen, L., and Graybeal, B. (2012a). “Modeling structural performance of Eng., 10.1061/(ASCE)BE.1943-5592.0000173, 422–430.
second-generation ultrahigh-performance concrete pi-beams.” J. Bridge Russell, H. G. (2009). Adjacent precast concrete box-beam bridges:
Eng., 10.1061/(ASCE)BE.1943-5592.0000301, 634–643. Connection details, Transportation Research Board, Washington, DC.
Chen, L., and Graybeal, B. (2012b). “Modeling structural performance of Russell, H. G. (2011). “Adjacent precast concrete box-beam bridges: State
ultrahigh performance concrete I-beams.” J. Bridge Eng., 10.1061 of the practice.” PCI J., 56(1), 75–91.
/(ASCE)BE.1943-5592.0000395, 754–764. Russell, H. G., and Graybeal, B. A. (2013). “Ultra-high performance concrete:
Chen, L., and Graybeal, B. A. (2010). “Finite element analysis of ultra-high A state-of-the-art report for the bridge community.” Rep. No. FHWA-
performance concrete: Modeling structural performance of an HRT-13-060, Federal Highway Administration, Washington, DC.
AASHTO Type II beam and a 2nd generation pi-beam.” U.S. Dept. of Santos, P., and Júlio, E. N. (2012). “A state-of-the-art review on shear-
friction.” Eng. Struct., 45(Dec), 435–448.
Transportation, Rep. No. FHWA-HRT-11-020, National Technical
Santos, P., Júlio, E. N., and Silva, V. (2007). “Correlation between con-
Information Service Accession No. PB2011-100864, Federal Highway
crete-to-concrete bond strength and the roughness of the substrate sur-
Administration, Washington, DC.
face.” Constr. Build. Mater., 21(8), 1688–1695.
Dias-da-Costa, D., Alfaiate, J., and Júlio, E. N. B. S. (2012). “FE modeling
Sargand, S. M., Walsh, K. K., Hussein, H. H., Al Rikabi, F. T., and
of the interfacial behaviour of composite concrete members.” Constr.
Steinberg, E. P. (2017). “Modeling the shear connection in adjacent
Build. Mater., 26(1), 233–243.
box-beam bridges with ultrahigh-performance concrete joints. II: Load
Fu, C., Pan, Z., and Ahmed, M. (2011). “Transverse posttensioning design
transfer mechanism.” J. Bridge Eng., 10.1061/(ASCE)BE.1943-5592
of adjacent precast solid multibeam bridges.” J. Perform. Constr. Facil.,
.0001071, 04017044.
10.1061/(ASCE)CF.1943-5509.0000147, 223–230. Steinberg, E., Huffman, J., Ubbing, J., and Giraldo-Londoño, O. (2013).
Grace, N., Ushijima, K., Baah, P., and Bebawy, M. (2013). “Flexural behav- “Finite element modeling of adjacent prestressed concrete box-beams.”
ior of a carbon fiber–reinforced polymer prestressed decked bulb T- Proc., PCI National Bridge Conf., Precast/Prestressed Concrete
beam bridge system.” J. Compos. Constr., 10.1061/(ASCE)CC.1943 Institute, Chicago.
-5614.0000345, 497–506. Steinberg, E., Semendary, A., and Walsh, K. (2015). “Adjacent precast
Grace, N. F., Jensen, E., Matsagar, V., Bebawy, M., Soliman, E., and box-beam bridges: Using UHPC longitudinal joints.” Constr. Specifier,
Hanson, J. (2008). “Use of unbonded CFCC for transverse post-tension- 68(8), 28–42.
ing of side-by-side box-beam bridges.” Research Rep. RC-1509, Steinberg, E., Ubbing, J., Giraldo-Londoño, O., and Semendary, A. (2014).
Michigan DOT, Lansing, MI. “Parametric analysis of adjacent prestressed concrete box-beams with
Graybeal, B. (2014). “Design and construction of field-cast UHPC connec- UHPC-dowel shear keys.” Proc., PCI National Bridge Conf., Precast
tions.” Rep. No. FHWA-HRT-14-084, Federal Highway Administration, /Prestressed Concrete Institute, Chicago.
Washington, DC. Ulku, E., Attanayake, U., and Aktan, H. M. (2010). “Rationally designed
Graybeal, B. A. (2006). “Material property characterization of ultra-high per- staged posttensioning to abate reflective cracking on side-by-side box-
formance concrete.” Rep. No. FHWA-HRT-06-103, Federal Highway beam bridge decks.” Transportation Research Record, 2172, 87–95.
Administration, Washington, DC. Yuan, J., and Graybeal, B. (2016). “Full-scale testing of shear key details
Harris, D. K., Carbonell, M. A., Gheitasi, A., Ahlborn, T. M., and Rush, for precast concrete box-beam bridges.” J. Bridge Eng., 10.1061
S. V. (2014). “The challenges related to interface bond characterization /(ASCE)BE.1943-5592.0000906, 04016043.
of ultra-high-performance concrete with implications for bridge rehabil- Zilch, K., and Reinecke, R. (2000). “Capacity of shear joints between high-
itation practices.” Adv. Civ. Eng. Mater., 4(2), 75–101. strength precast elements and normal-strength cast-in-place decks.”
Hussein, H., Walsh, K., Sargand, S., and Steinberg, E. (2016). “Interfacial Proc., PCI/FHWA/FIB Int. Symp. on High Performance Concrete,
properties of ultrahigh-performance concrete and high-strength concrete Orlando, FL, Precast/Prestressed Concrete Institute, Chicago.